Scalable Diffusion Models with Transformers
2023pp. 4172–4182
Citations Over TimeTop 1% of 2023 papers
Abstract
We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops—through increased transformer depth/width or increased number of input tokens—consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512×512 and 256×256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter.
Related Papers
- → Dissecting FLOPs Along Input Dimensions for GreenAI Cost Estimations(2022)17 cited
- → FLOPs-efficient filter pruning via transfer scale for neural network acceleration(2021)12 cited
- → ParameterNet: Parameters Are All You Need(2023)5 cited
- → Towards Flops-constrained Face Recognition(2019)
- → Chapter 6: Flop counts(2022)